Name: PyLadies San Francisco Virtual Workshop
Start: 2025-12-04T18:00:00-08:00
End: 2025-12-04T19:30:00-08:00

**Virtual Workshop: Advanced Safeguarding for AI Inference Pipelines**

OpenAI recently received a court order to retain output logs, highlighting that **AI privacy can no longer be an afterthought**, it must be **front and center when deploying AI workloads**.

However, privacy in AI is still evolving, with much of the focus currently on training-time privacy and far less attention paid to **privacy during inference**. As we begin scaling prototypes into production-ready systems, it is essential to analyze our workflows and **intentionally integrate privacy-preserving techniques**.

In this workshop, we will focus on practical use cases for building AI applications, explore what can go wrong, and experiment with various privacy-enhancing methods tailored to different use cases.

**In this workshop, you will learn:**
1\. The basics of LLMs and how production\-ready applications are built using them
2\. The various privacy risks that can arise when using LLMs
3\. A conceptual overview of privacy\-preserving techniques: Differential Privacy\, Federated Learning\, and Homomorphic Encryption
4\. How to build an AI pipeline that includes integrated privacy safeguards
The session will include live demos and self-paced exercises. Attendees should be comfortable
with Python

**Tentative Agenda**

6:00 pm: Intro
6:05 pm: Community Announcements
6:15 pm: Workshop
6:50 pm: Q&A
7:00 pm: Community Announcements
7:10 pm: Networking and Debrief
7:30 pm: End

Alla Barbalat

PyLadies of San Francisco

Python Software Foundation Meetup Pro Network

Technology

Software Development

New Technology

Internet Professionals

Web Technology

Web Development

Web Design

Computer Programming

Python

Women in Technology

Django

Education & Technology

Technology Professionals

Social Coding

I’m Aishwarya, an AI Engineer at Prediction Guard. I work at the intersection of privacy, safety, and large language models (LLMs), with a focus on designing systems that are practical, grounded in real-world use, and safe.

My work spans building LLM pipelines with safety risks in mind, as well as fine-tuning open models for downstream tasks. Currently, I’m focused on evaluating LLMs across multiple dimensions for enterprise use cases—addressing pre-deployment risks, deployment-time issues, and online risk monitoring.

I’ve contributed to research in this space and am part of MLCommons, a community-driven AI safety initiative. I’ve also explored a range of privacy-preserving techniques, including differential privacy and federated learning.



PyLadies of San Francisco

PyLadies San Francisco Virtual Workshop

PyLadies of San Francisco

Details

Members are also interested in